Forward Looking Infrared Target Matching Algorithm Based on Depth Learning and Matrix Double Transformation

摘要

Targeting at the human target detection in infrared sequence images, the extraction method of feature region based on feature points is adopted. The depth-learning algorithm is firstly used to extract the feature points rapidly. Based on the feature points extracted by matrix double transformation, LBP algorithm is used to extract the feature region. After acquiring the feature region (ROI region) interested, feature extraction of wavelet entropy based on discrete wavelet transformation is conducted for ROI region. Then ROI region is classified through compound classification method.